“AI doctor” is on the job, can we trust it?

Late at night, in front of the smart terminal in the emergency department of Beijing Children's Hospital, a new mother scanned her child's medical report. A few seconds later, the diagnosis and treatment recommendations on the screen were completely consistent with the diagnosis made by the doctor on duty half an hour ago...

Recently, the AI ​​application boom triggered by DeepSeek has swept the medical field, and the topic of smart medical care has once again sparked heated discussions. Recently, the "AI pediatrician" of Beijing Children's Hospital officially took office, making people's understanding of smart medical care more concrete.

When AI “puts on” a white coat, what kind of “chemical reaction” will occur in seeing a doctor?

Previously, Mr. Xu from Linyi, Shandong uploaded his child's blood test report to DeepSeek, and the diagnosis he got was basically consistent with that given by the doctor. Similar cases have begun to appear frequently on the Internet.

"The reason why AI can accurately diagnose diseases is due to its powerful data processing and analysis capabilities." Li Bohan, associate professor at the School of Artificial Intelligence at Nanjing University of Aeronautics and Astronautics, said that through deep learning of patients' medical history, symptoms, imaging data and other data, AI can identify complex patterns and associations between diseases and various characteristics, thereby providing a basis for diagnosis.

In terms of medical image analysis, AI has an extraordinary diagnostic accuracy. Not long ago, Zhejiang University launched an AI pathology assistant that integrates visual and language models. It can lock cancer lesions in pathology images within 1 to 3 seconds and has achieved an accuracy of more than 95% in the diagnosis of various cancers.

Smart healthcare not only brings accurate diagnosis, but also has the ability to quickly complete medical history collection, initial symptom screening, and improve the efficiency of doctors' consultations; it can assist in the diagnosis and treatment of difficult diseases and break through the limitations of doctors' memory; it can provide patients with personalized treatment plans and help doctors develop precise treatment plans, etc. According to statistics, the AI ​​triage system alone can save 26 million hours of ineffective waiting time for the country's top three hospitals each year.

When “AI doctors” start to work in large numbers, should we trust them?

"Although AI performs well in diagnosis, the risk of misdiagnosis still exists." Han Guojing, deputy director of the Respiratory Department of the First Medical Center of the General Hospital of the People's Liberation Army, reminded that the operation of AI depends on accurate and standardized information input. If there is bias in the data or defects in the algorithm, it may lead to misdiagnosis.

More importantly, AI cannot currently replace the clinical experience and intuition of human doctors. For some complex cases, the doctor's experience is still indispensable. "AI is very good at processing standardized medical data, but it is difficult to capture unstructured information such as a patient's hesitant sigh or cold sweat on the forehead," Han Guojing emphasized.

Research data shows that the current misdiagnosis rate of AI is about 2%-5%, mainly concentrated in the fields of mental illness and complex chronic diseases.

The "medical practice" qualification currently obtained by AI does not include the ability to sit for consultations independently. "When faced with the consultation of the 'AI doctor', we should give it a chance. We don't have to rely entirely on the diagnosis results. We can compare them with the diagnosis results of doctors and make a comprehensive judgment based on our own actual situation," said Han Guojing.

"AI Doctor" will be a good doctor's assistant, not a replacement. "AI can be an encyclopedia in a stethoscope, but the hand holding the stethoscope must have human warmth," said Li Bohan.

(Popular Science Times reporter Chen Jie)